Thread regarding IBM layoffs

There is nothing to divest when it comes to power. The brand of worthless. Power 10 Is 18 months away from availability, and it will cost 2x for comparable performance from x86.

I copied this from a thread posted earlier this week. Can anyone in Austin comment on this idea of “willingness to divest non-aligned parts of IBM”? It sure looks like Power (scale out, open, AI, etc etc) is being set up to get divested.

https://finance.yahoo.com/quotes/ibm,orcl,hpq,hpe,rht,msft,intc,emc/

Keep your eye on this part. It’s the next shoe to drop

“He also indicated his willingness to divest parts of the company not aligned with the company’s new focus”

I would argue that Power Management shares the blame. There was a constant struggle between the teams for resources. When the split happened the Power team tried to hoard the people they deemed the best and shipped the problem people to the other hallway. We were instructed not to help the OpenPower team with anything. That entire management chain up through JA is a disaster and their top priority seems to be self preservation at all costs.

@2srk

1. What makes you think that ibm can adapt a design cycle of a year down from 3 when they couldn't even design properly in 3.

2. In the arm market while the licencee and licencor aren't fighting as you said, there is a lot of fight amongst licensors - startups with 50m are really fast in design and can achieve in year what IBM couldn't or possibly cant in 3.

Qualcomm (major chip company amongst others) already went that route and failed. Even though Q isnt like apple but they were hell of a lot better than ibm.
If ibm goes into ARM, the market competition will k–l it in a year just like cloud wars.
Best thing is to cut the loss and sell power or whatever parts (ip etc.) while they can and move on.

Hey Sysygy, thanks for the good reads.

Sysygy here one last time on this topic:

If IBM wanted to do something really bold and disruptive in the datacenter server market, now would would be a good time to decide that "Power 10" should actually be "ARM 8". Adopting ARM 8 would instantly kickstart all of the things that Open Power fantasized of achieving while avoiding the traps that k–led Open Power. It's an openly licensed, broadly supported, mature technology whose licensor does not compete with its own licensees. So that's a lot of "plus ones" on the demand side.

It also gets IBM out of the business of developing and supporting a proprietary chip architecture with the resulting massive expense, slow development cycles, limited range of available systems, constant struggle to get software vendors to port when they know that their opportunity depends on IBM, etc. Who else is still trying to make a mainstream business out of a proprietary chip? Sun, DEC, HP: all long gone. It feels like the entire concept is twenty years out of date. So that's a lot of "plus ones" on the supply side.

Now the obvious question is: if that's such a good idea, why hasn't ARM taken off in the datacenter already? And my answer is: lack of standards to drive an ecosystem. While ARM8 is (kinda) standardized as a chip design, there is no standard "systems architecture" above the level of the instruction set, like there is for Intel. When you look at existing ARM servers, each one has its own memory architecture, I/O architecture, even proprietary/niche network interconnects - unlike Intel servers, where I can buy a motherboard from one place, memory from another, interface cards from another, throw them in a case and know the system will work. The ARM market today is lot like the personal computer market before IBM stepped in: lots of computers ran on the same processor, but no two were truly alike.

IBM is one of very few companies with the ability to drive that systems architecture standard. We can port our own essential software to the system (DB2, etc.) to kickstart it. We can direct Red Hat to make RHEL for ARM a "first class" offering, standardized for the common systems architecture. And let's face it, who knows more about engineering enterprise-quality systems around a RISC architecture than the existing Power engineering team?

Maybe IBM can do for the ARM market what we did for the PC market ... only this time with a better end game.

Enterprise Power has yet another problem besides Intel shooting at it. Enterprise power every three years obsoletes approx 1/3 of its install base to move to scale out power.(Power 10 will nuke P9) The problem is scale out power is far less profitable, and Enterprise power is not growing, but rather shrinking. The entire Enterprise market is under assault via Cloud, and Virtual server (see tidalscale as an example). Most cloud companies have funded virtual server initiatives as scale out is far far cheaper. Finally the economics of Power are below replacement costs. Approx 5-7% of power sales are Enterprise class boxes with 93-95% being scale out. If you can’t grow scale out, you are in big trouble as manufacturing costs will overtake your profit. I believe the bean counters at IBM have looked at the economics, and decided to shift their focus into the SW side of the house, and to sell the HW side of the house. I expect a global foundries type of deal to be reached with the potential buyer

@1lpw "Open Power Hot Take" guy here (I've adopted the pseudonym Sysygy so I can find my own posts again...)

I don't have nearly such a strong "take" on Power as an enterprise play as I do on the inevitably of Open Power's fate.

I do think that "letting go" of academic/research HPC is a good decision. Those opportunities are high prestige, low revenue, and very high opportunity cost in the sense that they dominate the R&D at the expense of all other opportunities. Coral is the prime example. Now, some people will say that the engineering breakthroughs that a Coral drives trickle down to the mainstream, but that's like trying to justify the Apollo program for the technology spinoffs... like aluminum foil and Tang. We did Apollo "because it's hard"; if the goal was aluminum foil, we didn't need to spend 5% of GDP for a decade to get it!

Focused on the right target markets, like AI for enterprises, Power could tell a pretty damn good story. ADAS is a great use case (apart from the massive recession in the car industry due to Covid-19... but it'll come back). So is anything to do with pharma, like d–g analysis: several Power supercomputers like Summit have been pivotal in e.g. d–g screening for potential Covid-19 vaccines. I'm hoping that pharma companies will pay attention to that and realize that there is the potential to do the same in their regular business.

But... the Power management team has to realize, these are good enterprise markets but not enterprise huge markets. Power is never going to be a mass market product that displaces Intel for everyday, mundane datacenter tasks like webserving or CRM or Accounts Receivable/Payable. Mid-range Intel boxes are perfectly "good enough" for these tasks, Power is never going to be able to match the economics for such tasks; it's enterprise opportunity is in places where Intel servers strain to reach, and where "the most performance per square foot per dollar" is what counts.

By the way, I call this misplaced belief that Power can extend down into the mainstream datacenter workload "delusions of adequacy" :-)

Open power was a decent idea. Power 9 was first to market with gen4 PCI. And Open CPAI was technology ahead of its time.

There is plenty of enthusiasm out there for an alternative to x86/ intel. Google was a big supporter in the beginning. So was the Chinese government.Both of those "mega customers" bent over backwards to make it successful.

But even with some decent technology advantages, the incompetent management of the Open POWER movement was ill equipped to deal with Intel, AMD, but most importantly the political realities inside IBM.

For the person who wrote the last comment, that was well written , what is your opinion on the future of power now that it will return to only focusing on enterprise

Granted, the internal dynamics were always against Open Power - but then, anybody with any knowledge of IBM history could have predicted that.

Without that, would Open Power have had a chance? No.

Open Power was always a fundamentally bad idea for many reasons. There was simply no way to create the Open Power "ecosystem" that IBM wanted in order to drive comparability with Intel's universality, _as long as IBM was also in the business of selling Power systems_. And critically, IBM made a lot more money from systems than from licensing. If IBM made its Open Power "partners" successful, they would inevitably have succeeded at the expense of IBM Power; and if they failed, it would still have been at the expense of IBM Power's margins. To put it another way: if IBM equipped Open Power licensees for success, the Power business ceases to be attractive to IBM; if it handicapped it for failure, it ceases to be attractive to the licensees.*

Open Power only makes sense if Open Power is actually competing against Intel, and therefore can grow every partner's revenue (at the expense of Intel vendors). But the reality always was that Open Power was competing against Power, fighting over the crumbs from the same small slice of pie.

Was there a smarter way to do this? Maybe, but it would have required IBM to be willing to give up making Power systems itself and sell just the processors or motherboards, like Intel**, probably with a "second source" licensee as many buyers demand that. Or even get out of the processor hardware business entirely and just license the designs, like ARM. Maybe, just maybe, that would have allowed Power to become the defacto standard for the high end, e.g. in HPC - although it was still never going to replace Intel in the commodity server market. But IBM was not capable of such a bold move, because "shifting tin" generates revenue, and revenue, however unprofitable, is how most of IBM middle management is judged.

If all this is true, you might ask, why would anybody ever take a license for Open Power in the first place? Frankly, I think that several companies did it as a "favor" to IBM, expecting similar "consideration" in some other area in return, but never expecting much business from Open Power. (This attitude is not uncommon in Europe or east Asia.) Some did it because they were promised some sort of territorial exclusivity – but that just moved money from IBM’s left pocket to it’s right pocket, it didn’t grow the business as a whole. And some did it because, to put it bluntly, a lot of companies in our industry are impressed with great engineering but lack an understanding of what drives complex markets.

*You can see this same mistake being repeated all over our industry. For example, ever since DDN bought Lustre, other storage vendors have been looking for alternatives to Lustre.

**Yes, Intel has dabbled from time to time in own-brand servers, but always briefly, and never successfully.

Open power was a long shot but worth an attempt if ibm had any chance of growing power unfortunately they never had the funding and support to get it moving. As far as managers, I know atleast one FLM open power was let go with his team

Afaik Open power has not generated any revenue till date.

I agree that managers should be laid off too, something that never happens at ibm.

With US China spat on semiconductor, OpenPower will soon be deemed non-compliance as it enable many key technologies to China. I believe this is the last straw for the camel.

Power is done. It had a good run, and still is superior technology when you need speed or really big in memory data. Unfortunately most folks don’t need that, in that Intel is good enough. Couple that with the lack of SW/ISV solution investment, and you end up where we are at today. IBM started down the Invest in Power path when they abandoned Intel, but quickly lost their way to the LINUX shiny object and Power withered on the vine. Most customers just say “Intel is good enough” and that’s the end of that. Does anyone out there think IBM or anyone is going to stroke a check for superior technology when power has entered the “replacement” era of its life span. The only shot IBM has is China and their wish for a unique technology. Hello Lenovo

Open power never had a chance. The engineers are gone but the managers who oversaw and caused the failure remain to fail again at something else.

The sad thing they are still pretending nothing has changed. No one has said anything to the partners.

That was a mess, lots of talk, no funding, never made the hires they said they would. I agree it was a political fight over protecting their own and it was decisions by committee where people in charge of funding did not stand up and take ownership.

I am wondering if one of these partner companies especially those in China would buy Power or by a percentage, whatever is allowed politically. I can’t imagine a us company wanting power unless we give it away like we did our chip business.

Ideally sell it and have them make servers for our existing customers. Not sure what would happen if we just k–led like oracle did with sparc, what do we lose if we dump it, we are not competitive and future roadmap puts us further behind. Something is up, they have been hiding their cards from us and this drop of open power is the first step of something bigger. They brought a business lady from India to do the slashing, no technical background but she has business experiences with what is talking place now. Are leaders were pushed out, some retired and some moved to amd.